最小一乘法WiFi匹配定位之研究

由於WiFi定位常使用的KNN演算法，無法對訊號資料有進一步的分析，找出不良的訊號觀測量。因此，本文嘗詴利用具有穩健性的最小一乘法處理WiFi匹配定位（簡稱LAD），以獲得較精確的WiFi定位成果。實驗資料，包含模擬資料、室內及室外資料等WiFi資料。首先，以實驗資料測詴LAD精度，並與最小二乘法WiFi匹配定位（簡稱LS）及KNN定位精度比較。然後，再將粗差加入實驗資料，測詴LAD與LS的穩健性及偵錯能力。根據實驗成果顯示：（1）在定位精度方面，通常LAD優於LS，例如LS與LAD的室內定位資料，分別為2.41與2.20公尺。（2）在穩健性方面，於RSSI建模階段，LAD有很好的穩健性，LS則無；於定位解算階段，當粗差數量小於或等於2時，LAD有好的穩健性，LS則無。（3）在偵錯能力方面，於RSSI建模階段，LAD有很好的偵錯能力，可以從參考點的RSSI殘差判斷粗差所在；於定位解算階段，當粗差數量小於或等於2時，LAD有好的偵錯能力。

關鍵字

WiFi匹配定位；最小一乘法；最小二乘法；穩健性

並列摘要

Since the KNN algorithm of common WiFi positioning can't find the outliers in the WiFi signal. Hence, this research tries to use the robustness of Least Absolute Deviation to deal with WiFi matching positioning (short name, LAD) to overcome the sensibility of WiFi signal strength, expecting to make the result of WiFi positioning more reliable. Test data sets include simulated data, indoor data and outdoor data. At first, test data sets are used to study the accuracies of LAD, Least Squares WiFi matching positioning (short name, LS) and KNN. Then, the outliers are added to the test data sets. The robustness and the outlier detection ability of LAD and LS are tested accordingly. According to the test results, there are some findings. (1) Usually LAD has better position accuracies than LS. For example, the indoor positioning accuracies of LS and LAD are 2.41 and 2.20 m respectively. (2) LAD's robustness is better than LS. (3) LAD's outlier detection ability is better than LS. The outlier can be detected by the largest RSSI residual of reference points or check points.

並列關鍵字

WiFi Mathing Positioning ； Least Absolute Deviation ； Least Squares ； Robustness

主題瀏覽